📊 Logging & Monitoring AI Responses

Why Monitor AI Responses?

AI security doesn’t stop at real-time filtering. Effective logging & monitoring ensures:

✅ Full visibility into AI decisions → Track why prompts were accepted or blocked.
✅ Auditability & Compliance → Maintain a record of all AI interactions for security reviews.
✅ Continuous AI improvement → Use log data to fine-tune AI behavior & policies.

The AIandMe Firewall automatically logs every API request and response, enabling detailed post-analysis.

AIandMe Firewall Logs

Each firewall API call generates a unique log_id, which you can use to retrieve detailed assessments later.

📌 Example: Firewall Log Dashboard

📌 Example: Detailed Log Analysis

Retrieving Logs via API

You can fetch logs programmatically using the AIandMe API.

🔹 Request Format

Send a GET request with the log ID to retrieve the full AI analysis:

curl -X GET "https://your-firewall-endpoint/logs/{log_id}" \
     -H "Authorization: Bearer your-api-key"

🔹 Python Example

Here’s how to retrieve AIandMe Firewall logs using Python:

import requests
 
log_id = "your-log-id"
api_url = f"https://your-firewall-endpoint/logs/{log_id}"
headers = {"Authorization": "Bearer your-api-key"}
 
response = requests.get(api_url, headers=headers)
 
if response.status_code == 200:
    log_data = response.json()
    print("Log Details:", log_data)
else:
    print("Failed to retrieve log:", response.status_code)

Log Data Breakdown

The API returns structured JSON data, containing:

• status → pass or fail
• fail_category → off_topic, violation, or restriction
• assessment_rationale → Explanation of why the prompt passed or failed
• timestamp → When the request was processed

📌 Example Response:

{
  "log_id": "123456",
  "status": "fail",
  "fail_category": "restriction",
  "assessment_rationale": "User attempted to request sensitive data.",
  "timestamp": "2025-01-26T13:45:00Z"
}

Best Practices for AI Logging

• Regularly review logs → Identify patterns in AI failures.
• Use logs for model tuning → Improve accuracy based on failure cases.
• Enable alerts for critical failures → Automate responses to security violations.

💡 Need help? Check out FAQs or Join the AIandMe Community.